Detection and Modeling of Collective Emotions in Online Data

Basing upon emotionally annotated data from four different media (a set of blogs, BBC Forums, Digg portal and IRC channels) we demonstrate the collective character of affective phenomena in online communities. To test whether the emotions of a community member may influence the emotions of others, posts were grouped into clusters of messages with similar emotional valences. The frequency of long clusters was much higher than it would be if emotions occurred at random. Distributions for cluster lengths can be explained by preferential processes because conditional probabilities for consecutive messages grow as a power law with cluster length. Values of characteristic exponent describing this growth correspond to strength of affective attraction for various types of emotions. It is interesting that minor emotions display larger clustering effects, i.e. they interact stronger in a given community. We demonstrate also that our model of emotional clustering leads to emergence of persistent mono-emotional threads when the emotional cluster reaches a critical size. Such ordered patterns have been found in the Blog06 dataset although their number is limited by fluctuations and sentiment classification errors.